Connection limits
The relay enforces connection caps and rate limits to keep the service fair and stable. Exact values depend on your subscription/entitlement and may change; the mechanisms are:
Concurrent connection caps
| Scope | What it bounds |
|---|---|
| Per account | Total simultaneous connections across all your keys. |
| Per key | Simultaneous connections using a single key. |
| Per IP | Simultaneous connections from one source IP. |
When a cap is reached, additional connections are refused until you close some.
Connection-rate limit
There is a limit on how often you may open new connections (per IP and per account). Reconnect with a small backoff + jitter rather than hammering — a tight reconnect loop will be rate-limited.
Message hygiene
- Keep your consumer fast. If your client can't keep up, the relay drops your
backlog and sends a
gap— it does not block other clients. - One
symbol+channelper connection. Open multiple connections for multiple streams (within your caps).
Demo feed
The public demo feed is separately and more tightly limited: per-IP request limits on ticket minting, per-IP and global connection caps, a ~60s single-use lease, and a majors-only, depth-only scope. It is a preview, not a production feed.
Handling limits gracefully
- Reuse connections; don't reconnect per request.
- Back off with jitter on close/refusal.
- Spread load across keys/IPs only within the documented caps.